應用類神經模糊系統於季節風波浪之推算

Abstract

本文利用中央氣象局所提供之2001-2002年花蓮資料浮標資料，分析季節風與示性波浪之特性，以瞭解兩者之間的關係，進而應用類神經模糊系統，建立冬季與夏季之波浪推算模式，來推算示性波浪。 本文之模式經由不同網路輸入參數與訓練資料長度之比較，以24小時移動平均之風速作為網路輸入參數，且只需利用2001年1月之前15天與2001年6月之前15天，作為訓練資料，即可推算2002年冬季與夏季之示性波高。本模式所推估之2001年及2002年冬季示性波高RMS值分別為0.27m與0.34 m，而夏季推估之示性波高RMS值分別為0.3m與0.35m，此結果証實本模式在波浪預報上均有不錯之精度，未來可供海岸工程施工、港灣船隻作業參考應用。

This paper investigates the relationship between monsoon and corresponding waves from the data of year 2001-2002 at Hualien harbor and sets up an adaptive network-based fuzzy inference system (ANFIS) of calculating waves both in the summer and in the winter. Only 15-day wind data with 24-hour moving average provide sufficient inputs that are related to waves to train a valid ANFIS model. The proposed ANFIS model has high accuracy to calculate waves in the winter by a RMS of 0.27m for 2001 and 0.34m for 2002, respectively. For summer waves, the proposed model gives a slightly worse calculation than for winter waves by a RMS of 0.30m for 2001 and 0.35m for 2002, respectively. This model is applicable for fast simulating waves from observed wind data as a reference of marine construction and navigating.